On the Sensitivity of Satellite‐Derived Cloud Properties To Sensor Resolution and Broken Clouds

Abstract

The sensitivity of cloud properties derived from meteorological imagers to sensor resolution is investigated by a comparison of results obtained from ${\text{1×1\hspace{0.17em}km}}^2$ versus down‐sampled ${\text{3×3\hspace{0.17em}km}}^2$ resolution MODIS reflectances. Focus is put on deviations caused by the plane‐parallel albedo bias for completely overcast, single‐phase pixels (67% of cloudy pixels), and by broken clouds (20% of cloudy pixels). For pixels corresponding to overcast water clouds, a low bias of $\text{−5.6\%}$ in cloud optical thickness and $\text{−3.8\%}$ in liqud water path is observed. A method to significantly reduce these biases is described, if information on unresolved variability in reflectance is available. For broken clouds, retrieval results are found to be rather unreliable. Again, a significant improvement in accuracy is achieved, if information on the sub‐pixel cloud fraction is used in the retrieval.